Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1...

34
Predatory Short Selling Markus K. Brunnermeier 1 and Martin Oehmke 2 1 Princeton University, NBER, and CEPR; 2 Columbia University Abstract. Financial institutions may be vulnerable to predatory short selling. When the stock of a financial institution is shorted aggressively, leverage constraints imposed by short-term creditors can force the institution to liquidate long-term investments at fire sale prices. For financial institutions that are sufficiently close to their leverage constraints, predatory short selling equilibria co-exist with no-liquidation equilibria (the vulnerability region ), or may even be the unique equilibrium outcome (the doomed region ). Increased coordination among short sellers expands the doomed region, where liquidation is the unique equilibrium. Our model provides a potential justification for temporary restrictions of short selling for vulnerable institutions and can be used to assess recent empirical evidence on short-sale bans. JEL Classification: G01, G20, G21, G23, G28 For helpful comments, we thank Thierry Foucault (the editor), an anonymous referee, Charles Jones, Marco Pagano, Alejandro van der Ghote, and seminar participants at Columbia University.

Transcript of Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1...

Page 1: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

Predatory Short Selling

Markus K. Brunnermeier1 and Martin Oehmke2

1Princeton University, NBER, and CEPR; 2Columbia University

Abstract. Financial institutions may be vulnerable to predatory short selling. Whenthe stock of a financial institution is shorted aggressively, leverage constraints imposedby short-term creditors can force the institution to liquidate long-term investments at firesale prices. For financial institutions that are sufficiently close to their leverage constraints,predatory short selling equilibria co-exist with no-liquidation equilibria (the vulnerabilityregion), or may even be the unique equilibrium outcome (the doomed region). Increasedcoordination among short sellers expands the doomed region, where liquidation is theunique equilibrium. Our model provides a potential justification for temporary restrictionsof short selling for vulnerable institutions and can be used to assess recent empiricalevidence on short-sale bans.

JEL Classification: G01, G20, G21, G23, G28

For helpful comments, we thank Thierry Foucault (the editor), an anonymous referee,Charles Jones, Marco Pagano, Alejandro van der Ghote, and seminar participants atColumbia University.

Page 2: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 1

1. Introduction

The financial crisis of 2007-09 and the recent European sovereign debt crisis have led toa heated discussion on short selling financial stocks. For example, as financial stocks fellsharply in the spring, summer, and fall of 2008, a number of banks, most notably BearStearns, Lehman Brothers, and Morgan Stanley, blamed short sellers for their woes.1 Inresponse, the Securities and Exchange Commission (SEC) and a number of internationalfinancial regulators took measures against short selling; most significantly, some imposedtemporary restrictions on the short selling of financial stocks, some even on short sellingin general. In August 2011, when European banks were struggling because of losses due tothe European sovereign debt crisis, market regulators in France, Spain, Italy and Belgiumimposed temporary bans on short selling for some financial stocks.

On both occasions, the worry was that short selling was “predatory,” in the sense thatshort sellers were attempting to bring down fundamentally solvent financial institutions bytriggering self-fulfilling downward spirals. However, this line of argument is at odds withthe consensus view in economics, which—broadly speaking—says that there is nothingwrong with short selling. In fact, most economists would argue that short selling is avaluable activity—short sellers help enforce the law of one price, facilitate price discovery,and enhance liquidity. Moreover, short sale restrictions may lead to overvaluation andbubbles, and reduce the ability of investors to hedge exposures.2 In the light of thesefindings, is there any economic justification to impose restrictions on short selling thestock of financial institutions?

In this paper, we present a model of predatory short selling. We show that, even thoughshort selling activity is beneficial during “normal times,” at times of stress short sellerscan, in fact, destabilize financial institutions through predatory short sales. Predatoryshort selling can occur in our model because financial institutions are subject to leverageconstraints imposed by their short-term creditors and uninsured depositors. These lever-age constraints capture a first-order difference between financial institutions and regularcorporates: Relative to corporations that can match maturities of assets and liabilities, thebusiness model of a financial institution almost necessarily involves maturity and liquiditymismatch (see, e.g., Diamond and Dybvig (1983) and Brunnermeier, Gorton, and Krish-namurthy (2013)), which exposes financial institutions to sudden withdrawal of fundingin response declines in equity value. We show that, in the presence of such leverage con-straints, predatory short sellers that temporarily depress the stock price of a financial

1 See, for example, “Anatomy of the Morgan Stanley Panic,” Wall Street Journal, Novem-ber 24, 2008.2 For theoretical models on how short-sale constraints can lead to overvaluation, specu-lative trading, and bubbles, see, for example, Miller (1977), Harrison and Kreps (1978),Scheinkman and Xiong (2003), and Hong and Stein (2003). Diamond and Verrecchia (1987)show theoretically that a market with short-sale constraints incorporates information moreslowly than a market in which short sales are not restricted. Empirical evidence that shortsellers contribute to market efficiency and market quality can be found, among others, inDechow, Hutton, Meulbroek, and Sloan (2001), Desai, Krishnamurthy, and Kumar (2006),Bris, Goetzmann, and Zhu (2007) , Chang, Cheng, and Yu (2007), Boehmer, Jones, andZhang (2008), Saffi and Sigurdsson (2011), and Boehmer and Wu (2013). Short positionsare important hedging tools in a number of common trading strategies (e.g., hedgingoptions, convertible bonds, or market risk in long-short strategies).

Page 3: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

2 M. K. BRUNNERMEIER AND M. OEHMKE

institution can force the financial institution to sell long-term assets in order to repaydebt to satisfy their leverage constraint. When long-term assets have to be unwound at asufficient discount, the resulting losses for the financial institution allow predatory shortsellers to break even on their short positions.

Our model implies that financial institutions can be vulnerable to attacks from preda-tory short sellers when their balance sheets are weak. For financial institutions that aresufficiently close to their leverage constraints, predatory short selling equilibria co-existwith no-liquidation equilibria (the vulnerability region), or may even be the unique equi-librium outcome (the doomed region). In the vulnerability region there are two stableequilibria. In one equilibrium, no predatory short selling occurs. In that case, the financialinstitution does not violate its constraint and can hold its long-term investments untilmaturity. In the second equilibrium, however, predatory short selling causes the financialinstitution to violate its leverage constraint, leading to a complete liquidation of its long-term asset holdings. In the doomed region, there is a unique stable equilibrium in whichpredatory short sellers force the financial institution to liquidate its entire long-term assetholdings.

Comparing a regime with short selling to one with short-sale restrictions shows that,during “normal times” when financial institutions are well capitalized, the fundamentalvalue of the financial institution cannot be affected by the presence of predatory short sell-ers. In this region short sellers exclusively fulfill their beneficial roles of providing liquidityand preventing overvaluation and bubbles that may distort real investment decisions—confirming the consensus view that restricting short selling during normal times is likelyto have undesired consequences. However, this changes once a financial institution entersthe vulnerability region or the doomed region. Here, short sellers can force inefficient liqui-dation of the financial institutions’ long-term assets, such that restrictions on short sellingcan potentially be welfare-enhancing.

By highlighting the possibility of multiple equilibria, our model underlines the impor-tant role of coordination in short selling attacks. Specifically, adding a large short seller(or, equivalently, a mass of small short sellers that can coordinate their actions) to ourcompetitive benchmark model expands the doomed region, where a predatory short sellingattack and full liquidation is the unique (and inefficient) outcome. This contrasts sharplywith the situation where sellers are regular shareholders (rather than short sellers): In thiscase, adding a large shareholder (or a mass of coordinated small shareholders) increasesthe safety region, in which no liquidation is the unique equilibrium. Finally, we show thatwhen there is a large short seller and a potential large support buyer, the doomed regiondepends on the relative strength (or ability to coordinate) of the large short seller and thesupport buyer.

Overall, our results provide a potential justification for temporary short sale restrictionsfor financial institutions at times when their balance sheets are weak. These restrictionsshould be temporary and targeted specifically at weak financial institutions because well-capitalized financial institutions are not susceptible to predatory short selling attacks, suchthat the only effect of a ban on short selling for those institutions would be a reductionin liquidity and market quality of their stock. Moreover, because our results are driven bya constraint on market leverage that is imposed by short-term creditors, it highlights theparticular vulnerability of financial institutions—because of the maturity mismatch andliquidity mismatch inherent in their business models, financial institutions are subject tofinancial fragility in the form of creditor runs, which makes them vulnerable to predatoryshort sellers. Our model is less likely to apply to firms with more stable capital structures.

Page 4: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 3

Finally, our analysis has implications on the disclosure of short positions. By facilitatingcoordination among short sellers, full and timely disclosure of short positions may in factmake it easier for short sellers to prey on vulnerable financial institutions.

The empirical evidence on the recent short-sale bans in the U.S. and Europe unambigu-ously documents reductions in market liquidity and market quality as a result of shortsale bans. However, there is no strong empirical support for positive price effects of recentshort sale bans, perhaps the main motivation of these bans. Using daily international dataon recent short-sale bans around the world, Beber and Pagano (2013) document that theshort-sale bans implemented during the financial crisis of 2007-2009 led to reductions inmarket liquidity and slower price discovery. They also document that short-sale constraintsfailed to support stock prices, except potentially those of large U.S. financial stocks. Insimilar spirit, Boehmer, Jones, and Zhang (2013), using intraday data on the 2008 short-sale ban in the U.S., document a deterioration of liquidity and market quality in responseto the short-sale ban. Yet, their analysis also finds little evidence that short-sale bans sup-ported prices: Only the largest financial institutions had (permanent) positive abnormalreturns during 2008 short-sale ban, and Boehmer, Jones, and Zhang (2013) point out thatthe price effects of the short selling ban are hard to disentangle from the effects of thecontemporaneous Troubled Asset Relief Program (TARP).3

Our results may help interpret the existing empirical findings on the effect of shortselling bans, and may also be useful in the empirical design of future studies. First, whenlooking at price effects, our model suggests that financial institutions, rather than regularfirms, should be particularly affected by short sale bans. Second, in our model, the ability ofshort sellers to prey on financial institutions depends crucially on the financial condition ofthe financial institution. Hence, a second cross-sectional prediction of our model is that inassessing the price effects of short-sale restrictions, one should control for leverage, liquiditymismatch, or similar variables that measure financial fragility. The prediction of our modelis that it is vulnerable financial firms for which the price effects of short sale bans are likelylargest. Third, our model highlights the importance of taking into account the potentialmultiplicity of equilibria when interpreting the empirical evidence. For example, if investorsexpect that, with some probability, there is a switch to the dominated equilibrium inwhich the bank goes bankrupt, the elimination of the bad equilibrium through temporaryshort selling bans when financial institutions enter the vulnerability region may lead topermanent positive price effects. In addition to the cross-sectional predictions above, a

3 A number of other studies have investigated the effects of recent short-sale bans. Forexample, Marsh and Payne (2012) document a decrease in market quality in UK equitymarkets in response to the 2008 short selling ban. Using U.S. and European data, Lioui(2011) documents an increase in volatility in response to the 2008 short selling ban, but noeffect on the price skewness. Autore, Billingsley, and Kovacs (2011) document that duringthe 2008 short sale ban in the U.S., stocks with a larger decline in liquidity also havepoorer contemporaneous returns, consistent with the model of Amihud and Mendelson(1986). Harris, Namvar, and Phillips (2013) use a factor model to document price inflationin banned stocks as a result of the 2008 short selling ban in the U.S., particularly forfirms without traded options. Their results suggest that price effects were temporary forstocks with negative pre-ban performance and permanent for firms with positive pre-banperformance. Kolasinski, Reed, and Thornock (2013) document that variations in shortinterest had larger price effect during the shorting ban, consistent with an increase in theinformativeness of short sales in response to increased short sale restrictions.

Page 5: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

4 M. K. BRUNNERMEIER AND M. OEHMKE

novel prediction of our model is that the vulnerability of a financial institution to predatoryshort selling depends not only on its own balance sheet but also on the balance sheets (orfunding conditions) of its large shareholders.

At a theoretical level, the potential justification for restrictions on short selling providedby our model is similar to that given in the literature on feedback effects from stockprices to firms’ real decisions.4 For example, Goldstein and Gumbel (2008) provide anasymmetric information model, in which a feedback loop to real investment decisionsallows a short seller to make a profit even in the absence of fundamental information.Khanna and Mathews (2012) study the interaction between an uninformed speculator andan informed blockholder to a firm. They show that, under certain conditions, manipulationby an uninformed speculator is possible even in the presence of an informed blockholder,whose incentives are aligned with value maximization. A major difference between thesepapers and our analysis is the channel through which short selling can be profitable. Inboth Goldstein and Gumbel (2008) and Khanna and Mathews (2012), short sellers reduceprice informativeness, thereby inducing the firm (whose manager learns from prices) toinefficiently distort its (future) investments, which makes the short position profitable. Inour framework, price declines brought about by short sellers can trigger inefficient earlyliquidation of existing investments via the leverage constraint. In our view, this latterchannel is particularly relevant for the recent discussion on short-sale bans, which hascentered around financial institutions.A related feedback mechanism arises in Goldstein,Ozdenoren, and Yuan (2013). In their model, a provider of equity capital learns from thefirm’s stock price and provides less capital to the firm when he infers negative informationfrom the firm’s stock price.5

In terms of the focus on financial institutions, the paper closest to ours is Liu (2011). Liudevelops a two-stage global games model: In the first stage, short sellers take positions. Inthe second stage, creditors decide whether or not to roll over their debt. As in Goldstein andGumbel (2008), the presence of short sellers reduces price informativeness. The resultingincrease in uncertainty about fundamentals reduces the value of short-term debt claims(due to their concave payoff) and can induce creditors to run in the second stage. Hence,a major difference to our paper is how short-sale attacks work: In our framework it is thereduction in the market value of equity that makes a short selling attack possible. In Liu(2011), it is the increase in price uncertainty (and not the price reduction) that leads toa creditor run. Liu’s framework leads to policy prescriptions that are broadly in line withours. First, as in our model, his framework implies that banks with weak fundamentalsare prone to short-sale attacks. Second, Liu argues that more maturity mismatch makesshort-selling attacks more likely. This is consistent with our model, where one can interpretthe severity of the leverage constraint as a proxy for maturity mismatch (the more shortterm creditors the bank has, the more binding the run constraint).6

4 For a comprehensive survey of this literature, see Bond, Edmans, and Golstein (2012).5 Khanna and Sonti (2004) and Ozdenoren and Yuan (2008) provide models with exoge-nous feedback from financial markets to real outcomes.6 More broadly, our paper also relates to the literature on market manipulation, andpredatory trading. Allen and Gale (1992) provide a model in which a non-informedtrader can make a profit if investors think the manipulator may be an informed trader.Other papers that consider manipulative trading strategies include Allen and Gorton(1992), Benabou and Laroque (1992), Kumar and Seppi (1992), Gerard and Nanda (1993),Chakraborty and Yilmaz (2004), and Brunnermeier (2005). Brunnermeier and Pedersen

Page 6: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 5

The remainder of the paper is structured as follows. Section 2. gives a brief summaryof regulatory response to short selling during the financial crises of 2007-2009 and theEuropean sovereign debt crisis of 2011. Section 3. presents the model. Section 4. pro-vides a discussion of the model’s policy implications and empirical predictions. Section 5.concludes. All proofs are in the Appendix.

2. Recent regulatory response to short selling

As a result of the financial market turmoil in 2008, the SEC and a number of internationalfinancial market regulators put in effect a number of new rules regarding short selling. InJuly the SEC issued an emergency order banning so-called “naked” short selling7 in thesecurities of Fannie Mae, Freddie Mac, and primary dealers at commercial and investmentbanks. In total 18 stocks were included in the ban, which took effect on Monday July 21and was in effect until August 12.

On September 19 2008, the SEC banned all short selling of stocks of financial companies.This much broader ban initially included a total of 799 firms, and more firms were addedto this list over time. In a statement regarding the ban, SEC Chairman Christopher Coxsaid, “The Commission is committed to using every weapon in its arsenal to combatmarket manipulation that threatens investors and capital markets. The emergency ordertemporarily banning short selling of financial stocks will restore equilibrium to markets.This action, which would not be necessary in a well-functioning market, is temporary innature and part of the comprehensive set of steps being taken by the Federal Reserve, theTreasury, and the Congress.” This broad ban of all short selling in financial institutionswas initially set to expire on October 2, but was extended until Wednesday October 9,i.e., three days after the emergency legislation (the bailout package) was passed.

In addition to measures taken by the SEC, a number of international financial regulatorsalso acted in response to short selling. On September 21 2008, Australia temporarilybanned all forms of short selling, with only market makers in options markets allowed totake covered short positions to hedge. In Great Britain, the Financial Services Authority(FSA) enacted a moratorium on short selling of 29 financial institutions from September18 2008 until January 16 2009. Also Germany, Ireland, Switzerland and Canada bannedshort selling of some financial stocks, while France, the Netherlands and Belgium bannednaked short selling of financial companies.

International restrictions on short selling of financial stocks reappeared in 2011. InAugust of 2011, market regulators in France, Spain, Italy and Belgium imposed temporaryrestrictions on the short selling of certain financial stocks as European banks came underincreasing pressure as part of the sovereign debt crisis in Europe. For example, both Spainand Italy imposed a temporary bans on new short positions, or increases in existing shortpositions, for a number of financial shares. France temporarily restricted short selling for11 companies, including Axa, BNP Paribas and Credit Agricole.8 On August 26, France,

(2005) provide a model in which a predatory trader can exploit another trader’s need toliquidate.7 In a naked short-sale transaction, the short seller does not borrow the share beforeentering the short position.8 See Howard Mustoe and Jesse Westbrook, “Short Selling of Stocks Banned in France,Spain,” Bloomberg, August 12, 2011.

Page 7: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

6 M. K. BRUNNERMEIER AND M. OEHMKE

Italy and Spain extended their temporary bans on short selling until at least the end ofSeptember.

Of course, measures against short selling are not exclusive to these recent episodes. Inresponse to the market crash of 1929, the SEC enacted the uptick rule, which restrictstraders from selling short on a downtick. In 1940, legislation was passed that bannedmutual funds from short selling. Both of these restriction were in effect until 2007. Goingback even further in time, the UK banned short selling in the 1630s in response to theDutch tulip mania.

3. Model

We consider a simple model with three periods, t = 0, 1, 2. At t = 0, a financial institutionhas invested in X units of a long-term asset. The financial institution has also taken outdemandable debt with face value D0. We take both the initial position in the risky assetas well as the initial debt outstanding as given.

Most of our analysis focuses on the interim date t = 1. Seen from t = 1, the long-term asset is expected to pay off a deterministic amount R at t = 2. If needed, the long-term asset can be liquidated at t = 1, but early liquidation is subject to a discount; theliquidation value at t = 1 is given by δR, where δ < 1. Hence, early liquidation is inefficient.For simplicity, we assume that the financial institution holds no cash, but the model couldbe straightforwardly extended to allow for cash holdings.

Leverage Constraint. Key to our analysis is that the financial institution is subjectto a leverage constraint. Specifically, we assume that debt as a fraction of debt plus themarket value of equity cannot exceed a critical amount γ ∈ [0, 1]:

D

D +E≤ γ. (1)

This leverage constraint captures in a simple way a fundamental difference betweenfinancial institutions and regular corporations. Specifically, relative to corporations thatcan match maturities of assets and liabilities, the business model of a financial institu-tion almost necessarily involves maturity and liquidity mismatch (see, e.g., Diamond andDybvig (1983) and Brunnermeier, Gorton, and Krishnamurthy (2013)). Moreover, beyondthe maturity mismatch that is inherent in their business model, financial institutions mayhave an additional incentive to take on significant maturity mismatch because of collectivemoral hazard (Farhi and Tirole (2012)) or because their inability to commit to longer-termfinancing leads to a maturity rat race (Brunnermeier and Oehmke (2013)).

In the presence of maturity mismatch, the leverage constraint (1) emerges because unin-sured depositors and creditors of the bank withdraw their funding when the bank’s marketleverage exceeds γ at the interim date. While we do not model this formally, what we havein mind is that creditors use the market price of equity to update their expectations onthe bank’s prospects and refuse to roll over their loans whenever the financial institution’smarket leverage exceeds a threshold γ. One can thus think of the leverage constraint asa “run constraint,” in the sense that creditors run on the bank following negative sig-nals about the value of the bank’s equity relative to its outstanding debt obligations (asin models of fundamental bank runs, such as Postlewaite and Vives (1987), Jacklin andBhattacharya (1988), or Goldstein and Pauzner (2005)). This interpretation of the lever-age constraint also highlights the connection to the literature on feedback effects of asset

Page 8: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 7

prices—the constraint captures, in reduced form, the feedback that arises when providersof capital learn from prices (as, for example, in Goldstein, Ozdenoren, and Yuan (2013)).

We formulate the leverage constraint in terms of market leverage. However, what ulti-mately is important for the model is that—independent of its particular form—the leverageconstraint implies that when the financial institution’s market value of equity falls belowa certain value, the financial institution is forced to liquidate some of its long-term assetsin order to repay creditors. As we will show below, in certain circumstances this feedbackmechanism, triggered by stock price declines, can make the financial institution vulnerableto short sellers in the equity market.9

Equity Market. At date t = 1, the equity of the financial institution is traded in afinancial market. This financial market is populated by two types of investors, a compet-itive fringe of passive long-term investors and active traders which act as short sellers.More generally, one could think of the short sellers as arbitrageurs that can take bothlong or short positions. However, since the main results of our paper revolve around theeffects of short selling, we will refer to them as short sellers. We also assume that shortsellers start with a zero position in the financial institution’s equity. We discuss the caseof regular sellers and differences between selling and short selling in Section 3.3.

The long-term investors are competitive and offer demand schedules to the short sellers.The long-term investors are thus not active traders themselves, they simply form a residualdemand curve that short sellers can sell into. Upon observing the demand schedules offeredby the long-term investors, short sellers decide whether to take a position in the stock.Short sellers are competitive and thus make zero profits in equilibrium.

We focus on the interaction in the equity market at the intermediate period, t = 1.At t = 1, the two types of players, long-term investors and short sellers, interact in thefollowing way. Long-term investors choose the slope and intercept of a demand schedulethat they offer to the short sellers. We denote the intercept by P and the slope by λ.Formally the long-term investors’ action space is thus given by the pair (P , λ) ∈ R× R+.Note that by assumption λ > 0, i.e., the residual demand curve for the stock is downwardsloping. However, as we will argue below, the slope of the demand curve can be arbitrarilysmall.10 Upon observing the demand schedules offered to them, the short sellers decidehow much of the stock they want to sell short. Their action space is thus the size oftheir short position, S ∈ R.11 Given these ingredients, we can write price of the financialinstitution’s equity at t = 1 as

P = P − λS. (2)

9 This particular vulnerability of financial institution’s is echoed in the SEC’s justificationof the 2008 short selling ban, which highlights the potential loss of confidence of tradingcounterparties in response to short selling.10 There are a number of ways to justify a downward-sloping demand curve. For example,our assumption may capture in reduced form that long-term investors are risk averse andneed to be compensated for risk that hey hold in equilibrium. The downward slopingdemand curve may also the be the result of information asymmetries (as in Kyle (1985))that are not modeled explicitly here.11 While we focus on short positions, we do not rule out long positions, which can betaken by picking a negative S. Hence, short sellers in this model are not short sellersby assumption—rather, they take short positions to exploit the financial institution’sconstraint, which is not possible by taking long positions.

Page 9: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

8 M. K. BRUNNERMEIER AND M. OEHMKE

Equilibrium. The equilibrium amount of short selling will be determined by a zeroprofit condition, meaning that the stock price at t = 1, P = P − λS must be a rationalprediction of the value of equity at t = 2, when the long-term investment pays off andequity investors receive their payoff. Denoting the payoff to equity holders at date t = 2by P , the equilibrium condition is thus given by

P = P. (3)

Predatory Short Selling. We distinguish two types of short selling. In an equilibriumwith regular short selling, short sellers are active, but their only effect is to ensure that thestock price coincides with the given fundamental value of the financial institution’s equity.In other words, regular short selling ensures that the price is right, but does not affect thefundamental value of the firm. In an equilibrium with predatory short selling, on the otherhand, the act of short selling reduces the fundamental value of the financial institution—itforces early liquidation of long-term assets at a loss. Through this feedback mechanism,predatory short selling reduces the value of equity and thus becomes self fulfilling. Hence,while regular short selling ensures that prices are right for a given fundamental and is thusbeneficial, predatory short selling destroys fundamental value and is inefficient.

Note that, in our formulation, long-term investors act essentially as passive shareholders.Specifically, we rule out that current shareholders can meet the leverage constraint byrecapitalizing the bank. This assumption reflects that recapitalization via issuing newequity may be hard in the midst of a short selling attack (in addition to the more generalobservation that issuing equity may be costly because of the usual asymmetric informationconsiderations). Similarly, we rule out that the financial institution renegotiates its debt,for example through a debt for equity swap. This assumption seems reasonable giventhat a financial institution that faces dispersed short-term creditors (or depositors) willusually have a hard time renegotiating debt, because of the coordination issues inherentin renegotiating dispersed debt issues.12

3.1 BENCHMARK CASE WITHOUT LEVERAGE CONSTRAINT

We first solve a benchmark model without the leverage constraint. As we will see, inthis setup short sellers serve a role in ensuring that the financial institution’s equity isfairly priced (through regular short selling), but the short sellers’ actions do not have anyinfluence on the fundamental value of the financial institution. Hence, in the absence ofthe leverage constraint, predatory short selling cannot occur in equilibrium.

Lemma 1. When financial institutions are not subject to the leverage constraint (1),predatory short selling does not occur in equilibrium.

The intuition behind Lemma 1 is simple. Because absent the leverage constraint thefundamental payoff to equity holders is fixed at XR−D0, only regular short selling canoccur in equilibrium: short sellers may take a short position to ensure that the financial

12 Potentially, the coordination problems in renegotiating debt could be mitigated if thefinancial institution issued some amount reverse convertible debt that can be convertedinto equity when the leverage constraint is binding. However, recall that in the Diamondand Rajan (2001) model of banking, it is precisely the inability to renegotiate debt thatmakes financing with dispersed short-term efficient for financial institutions.

Page 10: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 9

institution’s equity is valued correctly in cases where the long-term investors pick anintercept that exceeds fundamental value, i.e., P > XR−D0. However, there is no wayfor short sellers to affect the fundamental value of the financial institution’s equity, whichmakes predatory short selling impossible.

Lemma 1 thus highlights the beneficial role of regular short selling. When equity is over-priced relative to fundamental value, the ability to take short positions allows short sellers(or, more generally, arbitrageurs) to correct such overvaluation and make sure that equityis fairly priced. This is beneficial because overvaluation may lead to distorted investmentincentives and, ultimately, misallocation of resources. This beneficial role of regular shortselling is the main reason why unconditional short selling bans are undesirable.

For the remainder of the paper, we will focus on predatory short selling and put asidethe beneficial role of short selling illustrated in Lemma 1. To do this, we focus on thecase, in which the intercept chosen by long-term investors reflects the fundamental valueof equity in the absence of short selling. For example, if in the absence of short selling thefinancial institution does not have to sell any assets at the interim date, the long-terminvestors pick P = XR−D0. Focusing on the case implies that there is no role for regularshort sellers, since in the absence of leverage constraints short sellers would never take aposition. This assumption thus allows us to restrict our analysis to predatory short selling.However, as we discuss below, focusing on the case in which long-term investors set theintercept P equal to the fundamental value of equity in the absence of short selling isessentially without loss of generality: With slight adjustments the analysis generalizes toarbitrary P . The main difference between the two cases is that when P can deviate fromfundamental value, short sellers may also serve a beneficial role by reducing overpricing.

3.2 INTRODUCING THE LEVERAGE CONSTRAINT

We now introduce leverage constraint. Recall that the leverage constraint (1) requiresthe financial institution to keep leverage (defined as debt divided by debt plus the marketvalue of equity) below a critical level γ: D

D+E≤ γ. When the leverage constraint is violated

at date t = 1, the financial institution must repay some of its debt to reduce leverage andthus has to liquidate some of the long-term asset holdings.

Denote the number of units of the long-term asset the financial institution has to sellat t = 1 by ∆X(S), where S is the position taken by short sellers. If at t = 1 the financialinstitution sells ∆X(S) units of the long-term asset to repay debt, this leads to an equitypayout at time t = 2 of

P = max [XR−D0 − (1− δ)R∆X(S), 0] (4)

The reduction in equity value, (1− δ)R∆X(S), reflects the fact that the long-term assetcan only be sold at a discount. Using equation (4), we can thus rewrite the equilibriumcondition (3) as

P − λS = max [XR−D0 − (1− δ)R∆X(S), 0] . (5)

How much does the financial institution have to liquidate? In order to findpotential equilibria, we need to determine how much the financial institution needs toliquidate at t = 1. First note that when, given a short position of S, the leverage constraintis not violated, the financial institution does not have to liquidate any of its long-terminvestments. In this case ∆X(S) = 0. On the other hand, when the equity value at t = 1

Page 11: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

10 M. K. BRUNNERMEIER AND M. OEHMKE

(including the price effects of the short position S) is such that the constraint is violated,

D0

D0 + P − λS> γ, (6)

the financial institution has to sell ∆X(S) units of the long-term asset and repay debtin order to satisfy the constraint. The amount the financial institution has to liquidate isthen determined by the following condition:

D0 −∆X(S)δR

D0 −∆X(S)δR+ P − λS= γ. (7)

The numerator in (7) is the amount of debt remaining after liquidating ∆X(S) unitsof the long-term investment and thereby reducing outstanding debt by ∆X(S)δR. Thedenominator contains remaining debt D0 −∆X(S)δR plus the market value of equityP − λS. Solving (7) for ∆X(S) yields the following result:

Lemma 2. The amount of the long-term asset that the financial institution needs toliquidate under the leverage constraint (1) and in the presence of short sellers that take anaggregate short position S, is given by

∆X(S) =

{0 if D0

D0+P−λS≤ γ

min[(1−γ)D0−γ(P−λS)

Rδ(1−γ),X

]if D0

D0+P−λS> γ

(8)

Figure 1 illustrates what happens once we introduce the leverage constraint. In theillustration, absent short sales the leverage constraint is satisfied. However, a sufficientlylarge position by short sellers can force the financial institution to liquidate some of itslong-term asset holdings. These forced sales reduce the financial institution’s equity valueequity because the long-term asset has to be sold at a discount to fundamental value(when sold at t = 1 it yields δR rather than R). Hence, the fundamental value of thefinancial institution’s equity has a kink at the point where the leverage constraint becomesbinding and forces the financial institution to sell assets. To facilitate comparison to thebenchmark case discussed above, the dashed line indicates the fundamental value of equityin the absence of the leverage constraint.

Recall the equilibrium condition, P = P . This condition implies that potential equilibriaare intersections of the two lines in Figure 1, i.e., points where the price in the equitymarket at t = 1 rationally reflects the fundamental value of the equity of the financialinstitution at t = 2. Turn first to the top panel of Figure 1. We continue to assume thatlong-term investors choose the intercept P to be equal to fundamental value absent shortsales, i.e., P = XR−D0.

13 Because the two lines only intersect once, the only equilibriumremains the one without short selling—even though the short sellers can drive down thefundamental value of the financial institution by forcing it to liquidate some of its long-term investments, they invariably lose money in the process. This is the case whenever theliquidation value of the long-term asset, which is parameterized by δ, is sufficiently large.

13 As we show below, this assumption is not crucial in the sense that the equilibria areindependent of the particular choice of P and λ. We focus on the case P = XR−D0

because it allows us to focus exclusively on predatory short selling. The main differenceto the case P = XR−D0 is that now also the beneficial role of short selling (as discussedabove) emerges.

Page 12: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 11

In this case, the value destruction on response to a violation of the leverage constrainedis small, such that predatory short selling is not profitable. The unique equilibrium is theone where P = XR−D0. When P = XR−D0 this implies that the equilibrium amountof short selling is S = 0. More generally, when P = XR−D0, short selling can occurin equilibrium, but only in its beneficial role of ensuring that prices are equal to thefundamental XR−D0.

The bottom panel of Figure 1, on the other hand, shows that when δ is sufficientlysmall, predatory short selling can emerge. In this case, in addition to the equilibriumwithout short selling, two further equilibria emerge. Both of these additional equilibriainvolve predatory short selling: Short sellers cause a decrease in the financial institution’sequity value at t = 1, which forces the financial institution to liquidate long-term assetholdings to an extent that allows short sellers to break even. As is usually the case, themiddle equilibrium is unstable (such that a small perturbation would lead to migration toeither of the two stable equilibria).

Equilibrium prices are independent of λ and P . One convenient feature of ourmodel is that, as long as short sellers are not restricted in the number of shares they canshort, the equilibrium prices and the equilibrium amount of the long-term asset that thefinancial institution needs to liquidate are independent of the particular P and λ chosenby the long-term investors. Hence, while there are many equilibria involving differentcombinations of P , λ and S, these equilibria are isomorphic in terms of equilibrium pricesand liquidation quantities. One implication of this feature of the model is that whilesetting the intercept P equal to fundamental value absent short selling allows us to focusexclusively on predatory short selling, equilibrium prices and the existence of predatoryshort selling equilibria do not depend on this assumption.

Lemma 3. When short sellers are unconstrained in the size of the short position theytake, the equilibrium prices and the amount that has to be liquidated by the financialinstitution is independent of λ and P .

This independence result is illustrated in Figure 2 for the case in which the leverageconstraint is satisfied absent short selling. The top panel shows that when λ is decreasedfrom 0.75 (dashed line) to 0.6 (solid line), the equilibrium amount of short selling changes,but equilibrium prices remain the same. The bottom panel shows that when, in addition,also the intercept P is increased from 32 to 34, the equilibrium prices again remain un-changed. In this case, the equilibrium in which P = RX −D0 exhibits beneficial shortselling, while the other two equilibria exhibit predatory short selling.

Lemma 3 is convenient since it allows us to classify equilibria by looking only at equi-librium prices and the amount of the long-term asset that has to be liquidated by thefinancial institution.

Overview of Equilibria. We are now in a position to summarize the equilibria in theequity market at t = 1. In the proposition, we focus on the case where the long-term assetis relatively illiquid, δ < γ. This is the case depicted in the right panel of Figure 1. The(less interesting) case δ ≥ γ is discussed in the appendix.

Proposition 1. In the presence of the leverage constraint (1), when δ < γ we distinguishthree regions.

Page 13: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

12 M. K. BRUNNERMEIER AND M. OEHMKE

-10 0 10 20 30 40 50 60S

5

10

15

20

25

30

35

P, P�

P

P�

-10 0 10 20 30 40 50 60S

5

10

15

20

25

30

35

P, P�

P

P�

Fig. 1. Introducing the leverage constraint. When the leverage constraint isintroduced (in this figure, γ = 0.7), a sufficiently large short position can force thefinancial institution to liquidate some of its long-term asset holdings. In the toppanel, the loss to the financial institution from liquidating the long asset is notlarge enough to make a predatory short position profitable (δ = 0.75). The onlyequilibrium is the one in which no predatory short selling occurs. In the bottompanel, on the other hand, we see that when the losses from liquidating the long-termasset are large enough, two predatory short selling equilibria emerge in addition tothe equilibrium without predatory short selling (δ = 0.6). The middle equilibrium isunstable. The remaining parameters in this figure are: X = 10, R = 10,D = 68, λ =0.75

1. Safety region: When the financial institution is sufficiently well capitalized, R >D0δX

, there is a unique equilibrium in which the financial institution does not haveto liquidate any of its long-term holdings. No predatory short selling can occur,∆X(S) = 0, and P = XR−D0.

2. Vulnerability region: When D0γX

≤ R ≤ D0δX

, there are two stable equilibria andone unstable equilibrium.a. In one stable equilibrium, the financial institution does not liquidate any of its

long-term holdings, ∆X(S) = 0, and P = XR−D0.

Page 14: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 13

-10 0 10 20 30 40 50 60S

5

10

15

20

25

30

35

P, P�

-10 0 10 20 30 40 50 60S

5

10

15

20

25

30

35

P, P�

Fig. 2. Equilibrium prices are independent of P and λ. The top panelshows that when λ is decreased from 0.75 (dashed line) to 0.6 (solid line), theequilibrium amount of short selling changes, but equilibrium prices remain thesame. The bottom panel shows that when in addition also the intercept P is in-creased from 32 to 34, the equilibrium prices again remain unchanged. In this case,the equilibrium in which P = RX −D0 exhibits beneficial short selling, while theother two equilibria exhibit predatory short selling. The remaining parameters areR = 10,X = 10, δ = 0.6, γ = 0.7,D = 68.

b. In the other stable equilibrium, the financial institution is forced to liquidate itsentire holdings of the long-term asset, i.e., ∆X(S) = X and P = 0.

c. In the unstable equilibrium, the financial institution has to liquidate part of its

equity holdings, ∆X(S) = Xγ− D0

XRγ−δ

and P = 1−γγ−δ

(D0 − δXR)

3. Doomed region: When R < D0γX

, there is a unique stable equilibrium and an un-stable equilibrium.a. In the stable equilibrium, short sellers are active and the financial institution

liquidates its entire holdings of the long-term asset, ∆X(S) = X and P = 0.b. In the region D0

γX> R > D0

δX(1+γ)an unstable equilibrium exists, in which

short sellers are not active and the financial institution liquidates part

Page 15: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

14 M. K. BRUNNERMEIER AND M. OEHMKE

of the long-term asset holdings, ∆X(0) = XD0XR

−γ

δ−γ(1−δ)and P = XR−D0 −

1−δδ−γ(1−δ)

[D0 − γXR].

Figure 3 illustrates the equilibria as a function of XR, the fundamental value of thefinancial institution’s long-term asset holdings. As Proposition 1 points out, there arethree regions of interest. First, when R is sufficiently high, short sellers cannot profitablyforce the financial institution to liquidate long-term asset holdings. In this region, thefinancially institution is sufficiently well-capitalized, such that the only equilibrium is theone in which the financial institution holds its long-term investments until maturity. Werefer to this region as the safety region. In the safety region, short sellers solely fulfillthe beneficial function of correcting the equity value of the financial institution when thelong-term investors offer an intercept higher than the fundamental value of equity. Oneimportant implication from this region is that, when financial institutions are healthy oneshould not be concerned about predatory behavior by short sellers. Hence, our frameworkdoes not lend support to unconditional short selling bans.

However, when R drops sufficiently, there is a second region with multiple equilibria. Inthis region, the leverage constraint is satisfied if short sellers do not take predatory shortpositions. Hence, there still is an equilibrium without short selling and without liquida-tion by the financial institution. However, there are now also two equilibria in which shortsellers take predatory short positions and force the financial institution to liquidate someor all of its long-term asset holdings. In this region, the financial institution is vulnerableto predatory short selling, even though absent short selling the leverage constraint is notbinding. We thus refer to this region of multiple equilibria as the vulnerability region. Thisvulnerability region emerges only when δ < γ, i.e., when the long-term asset held by thefinancial institution is sufficiently illiquid. This highlights the importance of liquidity mis-match (illiquid long-term assets financed with short-term credit) in facilitating predatoryshort selling.

Finally, there is a third region with two equilibria, on stable and one unstable. In thestable equilibrium, short sellers are active and force the financial institution to liquidateits entire asset holdings, such that the equity value of the financial institution is given byP = 0. In the unstable equilibrium, short sellers are not active and the financial institutionliquidates part of its long-term asset holdings. Because in this region the unique stableequilibrium involves a complete liquidation of the financial institution, we refer to thisregion as the doomed region.

The effect of banning short selling. We are now in a position to compare outcomesunder a regime in which short selling is allowed and a regime in which short selling isprohibited. When short selling is restricted, the financial institution only needs to liqui-date at date t = 1 when the leverage constraint is violated absent predatory short sellers.Proposition 2 compares the two regimes, focusing on stable equilibria.

Proposition 2. Consider again the case δ < γ. The effect of banning short selling onequilibrium prices and the quantity of the long-term investment liquidated by the financialinstitution depends on the equilibrium region:

1. Safety region: When the financial institution is sufficiently well capitalized (R >D0δX

), equilibrium prices and the amount the financial institution needs to liquidatecoincide. In both cases, the unique equilibrium is P = XR−D0 and ∆X = 0.

Page 16: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 15

8 9 10 11 12 13 14XR

20

40

60

80

100

P

DoomedRegion

VulnerabilityRegion

SafetyRegion

short selling HstableL

short selling HunstableL

no short selling HunstableL

no short selling HstableL

Fig. 3. Overview of Equilibria. This plot shows the equilibrium equity valueof the financial institution as a function of R. For high values of R, there is aunique equilibrium without predatory short selling (safety region). Once R dropssufficiently low, there is a region with multiple equilibria (when γ > δ). In thisvulnerability region predatory short selling can emerge. The middle equilibrium isunstable. When R is so low that the leverage constraint binds in the absence of shortselling, there is a stable equilibrium with predatory short selling and an unstableequilibrium without predatory short selling (the doomed region). The parametervalues in this graph are X = 10, δ = 0.6, γ = 0.7,D = 68.

2. Vulnerability region: In the vulnerability region (D0γX

≤ R ≤ D0δX

), when short sell-ing is restricted no liquidation takes place and the unique equilibrium is given byP = XR−D0 and ∆X = 0. When short sellers are present, on the other hand,there is a second stable equilibrium, in which predatory short sellers force the fi-nancial institution to liquidate its entire long-term asset holdings: ∆X = X andP = 0.

3. Doomed region: In the doomed region (R < D0γX

), when short selling is re-stricted the financial institution liquidates part of its long-term asset holdings as

long as R > D0δX(1+γ)

. In this region, ∆X(0) = XD0XR

−γ

δ−γ(1−δ)and P = XR−D0 −

1−δδ−γ(1−δ)

[D0 − γXR]. When R ≤ D0δX(1+γ)

, the financial institution liquidates itsentire long-term asset holdings even when short selling is restricted, and P = 0.When short sellers are present, in the doomed region the financial institution al-ways liquidates its entire holdings and P = 0.

Figure 4 illustrates the main differences between a regime with short selling (solid line)and a regime in which short selling is restricted (dashed line). First, note that when short

Page 17: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

16 M. K. BRUNNERMEIER AND M. OEHMKE

8 9 10 11 12 13 14XR

20

40

60

80

100

P

Doomed RegionVulnerability

RegionSafetyRegion

Fig. 4. The effect of banning short selling. This figure compares equilibriawith and without short selling, focusing on stable equilibria. When short sellingis allowed (solid line), there are multiple equilibria once the financial institutionenters the vulnerability region. In one of the two stable equilibria, predatory shortsellers force the financial institution to liquidate its entire long-term asset holdings.In the doomed region, in the unique stable equilibrium short sellers always forcethe financial institution to unwind its entire asset holdings. When short selling isnot allowed (dashed line), the financial institution does not have to liquidate inthe vulnerability region and P = XR−D0. Moreover, in the doomed region, thefinancial institution only has to liquidate part of its long-term asset holdings whenshort selling is restricted (except when R is so low that the financial institution hasto liquidate everything even in the absence of short selling). The parameter valuesin this example are X = 10, δ = 0.6, γ = 0.7,D = 68.

sales are restricted, there is no vulnerability region—the financial institution only hasto liquidate some of its long-term asset holdings if the leverage constraint is violated inthe absence of temporary price movements caused by short sellers. When short sellingis allowed, on the other hand, the vulnerability region emerges and there is a secondequilibrium in which short sellers prey on the financial institution, forcing it to unwind itsentire long-term asset holdings. Hence, in this region predatory short sellers can force acollapse of the financial institution, even though the financial institution would be soundin the absence of short selling.

Second, when the leverage constraint is violated even in the absence of short selling, theamount the financial institution has to liquidate is (weakly) smaller when short selling isrestricted. This is the case because in the doomed region in the unique stable equilibriumshort sellers force the financial institution to liquidate its entire portfolio. When no shortsellers are present, on the other hand, the financial institution can in general satisfy theleverage constraint by selling only part of its long-term asset holdings, except when R

Page 18: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 17

drops so low that the financial institution enters a “death spiral” (i.e, it has to liquidateall long-term asset holdings even when no short sellers are present). In the figure, thishappens at the point where the dashed line meets the x-axis.

Of course, one caveat of the analysis above is that we have focused exclusively on thecase in which, absent short selling, the equity of the financial institution is priced correctly.This allowed us to focus exclusively on characterizing the conditions under which predatoryshort selling can occur. More generally, the potential welfare costs of predatory shortselling have to be weighed against the beneficial effects of regular short selling throughthe elimination of overvaluation and improvements in market quality and liquidity. Asdiscussed above, in the safety region predatory short selling cannot occur and the onlyeffect of short sellers is the elimination of mispricing. Clearly, in this region, a short-saleban is not desirable. In the vulnerability region, on the other hand, the costs of potentialpredatory short selling have to be weighed against the potential benefits from regular shortselling. The desirability of a potential short selling ban then depends on the relative sizeof these two effects. While formally our model does not deliver predictions on how largethese two effects may be in practice, it does provide some informal guidance. For example,if one believes that a financial institution is only temporarily in the vulnerability region,a short selling ban may prevent the collapse of the financial institution, while the costs oftemporary overvaluation of the financial institution’s equity might be moderate. Similarly,in the doomed region one would have to weigh the benefits of controlled deleveraging thatis possible in the absence of short sellers against the potential costs of overvaluation inthis region.

3.3 REGULAR SELLING VERSUS SHORT SELLING

Up to now our analysis has focused exclusively on short selling. In this section, we contrastour results to those that would obtain if we replaced short sellers with regular sellers(i.e., investors who have an initial endowment of shares in the financial institution). Thiswill sharpen the distinction between regular selling and short selling and highlight theimportant role important role of coordination.

The distinction between short selling and regular selling is particularly relevant in thevulnerability region where, as shown above, multiple equilibria Pareto-ranked equilibriaare possible: In the Pareto-dominant equilibrium, the financial institution does not haveto sell any of its long-term asset holdings and survives. In contrast, in the dominatedequilibrium, the financial institution is forced to liquidate all long-term asset holdingsand fails. The discussion in this section revolves around the following questions: Can thedominated equilibrium also emerge in a setting with regular sellers as opposed to shortsellers? If yes, is there reason to believe that it is more likely to emerge as a result of shortselling as opposed to regular selling?

Recall that in the competitive setup developed above, when all trades are executed atthe final market clearing price, short sellers are indifferent between the two equilibria thatare possible in the vulnerability region—they make zero profits in either. More generally, ifshort sellers can walk down the demand curve when establishing their short position (i.e.,when not all trades are executed at the final price) they strictly prefer the equilibrium inwhich they collectively prey on the financial institution. This contrasts with situation thatwould arise if short sellers were regular sellers: While a setup with regular sellers insteadof short sellers would lead to the same two equilibria in the vulnerability region, regular

Page 19: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

18 M. K. BRUNNERMEIER AND M. OEHMKE

sellers strictly prefer the equilibrium in which they hold on to their shares and do notsell. Hence, with regular sellers, the dominated equilibrium can only emerge as a result ofcoordination failure—existing shareholders sell because they expect everyone else to sell,comparable to the dominated equilibrium in Diamond and Dybvig (1983). While this, ofcourse, does not rule out the dominated equilibrium, it is a reasonable proposition thatthe dominated equilibrium is less likely to emerge through a pure coordination failure ofregular sellers than through a (weakly) profitable attack by predatory short sellers.

In addition, as soon as we depart from the competitive benchmark and allow for someamount of coordination among short sellers or shareholders, the equilibrium regions dif-fer depending on whether sellers are regular sellers or short sellers. Specifically, someamount of coordination among short sellers increases the doomed region where the uniqueequilibrium involves a complete liquidation of the financial institution. In contrast, someamount of coordination among regular sellers increases the safety region where the uniqueequilibrium involves no liquidation.

Formally, we model the degree of coordination by assuming that there is a large traderor, equivalently, a mass of small traders who can coordinate their actions. The largetrader (or coordinated traders) internalize that their trading decision moves the shareprice. The remaining traders form a competitive fringe and take prices as given. In theshort selling case, we assume that the large short seller can take a maximum short positionof SMAX. In the case of regular sellers, we assume that the large shareholder owns SMAX

shares in the financial institution, while the competitive fringe owns SMAXC shares. We

also assume that if both the large shareholder and the competitive fringe sell their shares,the share price drops to zero and the financial institution has to liquidate all of its long-term asset holdings. Formally, this assumption requires that in the regular selling caseSMAX + SMAX

C = S, where S is defined by P = P − λS = 0. Note that in both the regularand the short selling case SMAX proxies for the amount of coordination that is possible.

For simplicity and to reflect the role of large traders (such as George Soros) as firstmovers, we assume that the large trader moves first and that the competitive fringe movesafter the large trader’s order has been executed. However, as we describe in more detailbelow, the findings in Proposition 3 do not depend on the specific assumptions on moveand execution order. For example, we could alternatively assume that the large trader andthe competitive fringe submit their orders and are executed simultaneously, or that theexecution order is random and traders submit limit orders. Both of these alternative setupswould leave the equilibrium regions described in Proposition 3 unchanged (see footnotes16 and 17 for more details).

Consider first the case of short sellers. The large short seller moves first and choosesS ∈ [0, SMAX]. The large short seller’s trade is then executed at P (S) = P − λS. Then, thecompetitive fringe moves and chooses SC . The orders of the competitive fringe are executedat P (S + SC) = P − λ(S + SC). Whenever the maximum short position of the large shortseller, SMAX, is sufficiently large to make the short sale profitable irrespective of theactions of other short sellers, the unique equilibrium involves predatory short selling. Thisis the case whenever SMAX > S∗, where S∗ denotes the short position required to make ashort sale profitable (for a formal definition of S∗, see equation (A7) in the appendix). Thiscondition holds when the financial institution is sufficiently close to its leverage constraint.Hence, the presence of a large short seller expands the doomed region in which the uniqueequilibrium involves complete liquidation of the financial institution.

In contrast, in the case of regular sellers the presence of the large trader expands thesafety region in which the unique equilibrium involves no liquidation by the financial

Page 20: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 19

institution: This is the case when the blockholder’s decision not to sell his shares canensure that no coordination failure occurs, which is the case when SMAX > S − S∗: Giventhat the large shareholder does not sell, the competitive fringe cannot profitably coordinateto sell because P (SMAX

C ) < P (SMAXC ). The unique best response is thus SC = 0 and no

liquidation is the unique equilibrium.14 Solving for SMAX > S∗ and SMAX > S − S∗ interms of the parameters of the model yields the following proposition.

Proposition 3. Assume that there is a large trader or, equivalently, a mass of smalltraders that can coordinate there actions up to a maximum of SMAX shares.

1. Short selling: If traders who coordinate up to SMAX are short sellers,the doomed region (with a unique short selling equilibrium) expands to R ∈[0,min

[D0γX

+ γ−δ(1−δ)γX

λSMAX, D0δX

]).

2. Regular selling: If traders who coordinate up to SMAX are regular sellers,the safety region (with a unique no-liquidation equilibrium) expands to R ∈(max

[D0δX

− γ−δ(1−γ)δX

λSMAX, D0γX

],∞

).

The equilibrium regions in the presence of a large trader are illustrated in Figure 5. Thetop panel illustrates the expansion of the doomed region in the presence of a large shortseller. The bottom panel illustrates the expansion of the safety region in the presence ofa large shareholder in the regular selling case.

Proposition 3 shows that once a certain amount of coordination is possible, there is asharp difference between short selling and regular selling: While in both cases coordinationshrinks the parameter vulnerability region (the region with multiple equilibria), in theshort selling case this happens via an expansion of the doomed region (in which predatoryshort selling is the unique equilibrium), whereas in the case of regular sellers this happensthrough an expansion of the safety region (where no liquidation is the unique equilibrium).In the extreme, the vulnerability region vanishes completely. In the short selling case, thefinancial institution is then liquidated as soon as short sellers have the ability to force thefinancial institution to violate its leverage constraint (i.e., when R < D0

δX). In the case of

14 The role of the large short seller or blockholder discussed here is similar to the role oflarge players in the literature on currency crises. See, in particular, Corsetti, Pesenti, andRoubini (2002) and Corsetti, Dasgupta, Morris, and Shin (2004) for a setting in whichtraders face a binary decision on whether or not to attack a currency.

Page 21: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

20 M. K. BRUNNERMEIER AND M. OEHMKE

Fig. 5. Equilibrium regions under short selling and regular selling. Thefigure compares equilibrium regions under short selling and regular selling in thepresence of a large trader (or a mass of small traders who can coordinate theiractions) of size SMAX. The parameter region for which the unique equilibrium in-volves liquidation of the financial institution is larger under short selling than underregular selling. Conversely, the parameter region for which the unique equilibriuminvolves no liquidation by the financial institution is larger under regular sellingthan under short selling.

regular sellers, on the other hand, no liquidation occurs unless the financial institutionviolates its leverage constraint in the absence of short sellers (i.e., when R < D0

γX).15 16 17

15 One interesting implication of Proposition 3 is that the region in which the uniqueequilibrium involves predatory short selling depend on the slope of the demand curveλ, which we have taken as given here. This is the case when the position limit for theshort seller are in terms of the maximum number of shares that can be shorted, SMAX.Alternatively, if position limits are defined as “price impact” limits (which would be ofthe form SMAX/λ), the region in which predatory short selling is the unique equilibriumis independent of the slope of the demand curve λ, thereby recovering the irrelevanceproperty of Lemma 3.16 If instead of sequential orders and execution we were to assume that orders are sub-mitted and executed simultaneously, the resulting equilibrium regions would be identicalto those in Proposition 3. The main difference is that, in the simultaneous-move game,it is the threat of the large short seller that eliminates the no-liquidation equilibrium. As

Page 22: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 21

3.4 SUPPORT BUYING BY A LARGE TRADER

Next, we discuss the effect of adding investors who can step in to buy shares (recall thatup to now the long-term investors that form the residual demand curve were assumed tobe completely passive and thus never acted as active support buyers). To do this, considerthe case in which both a large short seller (or a mass of short sellers who can coordinate)and a large support buyer (or a mass of traders who can coordinate to purchase stock inthe financial institution) are present. We assume that the support buyer (this could, forexample, be a blockholder or another large trader with a vested interest in the financialinstitution) can buy up to BMAX additional shares to support the financial institution.For simplicity, we set the support buyer’s initial endowment in shares of the financialinstitution to zero. As before, the large short seller can short a maximum of SMAX shares.As in the previous subsection, we assume that the large traders (the short seller and thesupport buyer) trade first, followed by the competitive fringe.

In this case, the region in which predatory short selling is the unique equilibrium de-pends on the relative strength of the support buyer vis-a-vis the short seller. Specifically,starting from a conjectured no-liquidation equilibrium, the short seller’s maximum posi-tion SMAX must now be sufficiently large to make deviation profitable even if the supportbuyer purchases the maximum amount of shares BMAX. If this is the case, the uniqueequilibrium involves predatory short selling. Similar to Proposition 3, we can then char-acterize the regions in which the unique equilibrium involves predatory short selling asfollows:

Proposition 4. Assume that there is a large short seller or, equivalently, a mass ofsmall traders that can coordinate there actions up to a maximum of SMAX shares. Assumealso that there is a large support buyer (or a mass of small support buyers coordinate)who can purchase BMAX additional shares to support the share price of the financial in-stitution. Then the doomed region (with a unique short selling equilibrium) is given by

R ∈[0, D0

γX+ γ−δ

(1−δ)γXλ[SMAX −BMAX

]+).

before, when SMAX > S∗, the large short seller has a strictly profitable deviation from aconjectured no-liquidation strategy profile. However, the large short seller cannot be partof a zero-profit short selling equilibrium with S + SC = S, because from any such equi-librium he would have an incentive to slightly reduce the size of his short position andmake positive (instead of zero) profits. Hence, when SMAX > S∗ the unique equilibrium isa predatory short selling equilibrium in which the competitive fringe takes a short positionof SC = S in response to the threat of a short potions by the large short seller.17 Also a setup with random execution order in which traders can submit limit ordersleads to the same equilibrium regions. Because there is a one-to-one mapping betweenthe execution price and the order of execution, limit orders allow traders to effectivelycondition their sell orders on when they are executed. If the large trader is executed first,the analysis is identical to the one discussed in the text (the analysis in the text is alimiting case of the more general limit order setup: the probability that the large trader isexecuted first is one). In the case where the large trader is executed after the competitivefringe, the equilibrium regions remain the same but in the coordination failure equilibriumthe large shareholder may sell slightly less if he anticipates that his order will be executedafter the competitive fringe (and thereby at a lower price). See the appendix for moredetails.

Page 23: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

22 M. K. BRUNNERMEIER AND M. OEHMKE

Relative to Proposition 3, we thus see that the presence of a support buyer shrinksdoomed region, in which full liquidation is the unique equilibrium. This is because theboundary of the doomed region is now determined by the relative strength of the largeshort seller and the support buyer, as captured by SMAX −BMAX. Moreover, under theinterpretation that SMAX and BMAX reflect the extent to which multiple blockholders andmultiple short sellers can coordinate, Proposition 4 highlights how, in addition to theirsheer financial strength, the relative ability of blockholders and short sellers to coordinatebecomes an important element in determining which equilibrium obtains.

4. Discussion

While the simple model presented above does not provide a full welfare analysis of shortselling, our analysis generates a number of predictions that may help in devising a moredifferentiated regulatory approach to short selling. Moreover, the empirical predictions ofour model may be helpful in interpreting existing empirical evidence as well as providingguidance for the design of future empirical studies on short selling.

4.1 REGULATORY IMPLICATIONS

Vulnerability of financial institutions to short sales. One of the main implicationsof our model is to highlight the potential vulnerability of financial institutions to preda-tory short selling. In our model, predatory short selling can emerge because of a leverageconstraint that captures the run risk faced by financial institutions that inherently havesignificant maturity and liquidity mismatch. This run constraint allows short sellers to cap-italize on financial weakness by forcing an institution to liquidate long-term investments,leading to a reduction in fundamental value. This reduction in fundamental value, in turn,allows short sellers to break even on their positions. Hence, the potential unwillingnessof creditors to renew their funding leads to a fragility in financial institutions’ fundingstructures that can potentially be exploited by predatory short sellers. Firms with morestable capital structures, on the other hand, should be less susceptible to the predatorybehavior characterized in this paper.

Temporary short-sale bans. In terms of regulations that restrict short selling, ouranalysis implies that, while banning short selling during normal times is not desirable,it can make sense to restrict short selling of financial stocks temporarily when balancesheets across most financial institutions are weak: When banks are well-capitalized (andpredatory short selling does not occur in equilibrium), short sellers merely carry out theirbeneficial role of enforcing the law of one price, providing liquidity and incorporating in-formation into prices. Our model thus does not provide a justification for a general ban ofshort selling on the grounds of predatory behavior. However, when financial institutions’balance sheets are weak and they enter the vulnerability region, destabilizing predatoryshort selling can occur, leading to inefficient liquidation of long-term investments by vul-nerable institutions. Hence, while we want to stress that our paper does not provide a fullwelfare analysis of short selling, this result provides a potential justification for temporaryshort sales restrictions to curb predatory behavior (if one believes that the costs of po-tential predatory short selling outweigh potential overvaluation and reduction in liquidityduring a short selling ban).

Page 24: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 23

Disclosure of short positions. A major result of our analysis is the role of multiplic-ity of equilibria in the vulnerability region. Because there are two stable equilibria in thisregion, coordination among short sellers is crucial in determining which of the two equilib-ria we end up in. This has implications for the disclosure of short positions. In particular,in addition to recent short-sale restrictions, a number of regulators have enacted tougherdisclosure requirements for short positions. In the U.S., the SEC enacted a rule requiringinstitutional investors to publicly disclose their short positions on a weekly basis.18. In theUK, the FSA implemented a rule that requires that investors disclose on each day anyshort positions in excess of 0.25% of the ordinary share capital of financial companies atthe end of trading the previous day.19. Since November 1, 2012, EU regulation requiresshort sellers to report to regulators is they intend to short sell more than 0.2% of a com-pany’s tradable shares and to publicly report short positions that exceed 0.5% of tradableshares.20

However, our analysis indicates that public disclosure of short positions can, in fact, becounterproductive. In particular, requiring public disclosure of all short positions may infact facilitate coordination among predatory short sellers. When short sellers are requiredto publicly disclose positions, it may thus be more likely that we end up in the predatoryequilibrium when in the vulnerability region. One way to capture this formally in our modelis to consider an increase in the amount of shares up to which short sellers can coordinate,SMAX. It follows directly from Proposition 3 that such an increase in coordination enlargesthe doomed region, in which a predatory short selling attack is the unique equilibrium.This suggests that disclosure should either only be made to the regulator, or should bemade public only with sufficient time delay.

Other regulatory interventions. Up to now, we have limited our discussion of po-tential regulatory interventions to restricting short sales. In this section, we briefly discussother potential regulatory measures that may reduce the vulnerability of financial insti-tutions to predatory behavior by short sellers.

First, the leverage constraint arises because short-term creditors may withdraw fundingfrom the financial institution. This implies that interventions to increase the stability ofthe financial institution’s financing may be desirable. For example, such an interventioncould take the form of limiting liquidity mismatch, thus reducing financial fragility andincreasing the resilience of the financial institution against short selling attacks. Anotherpotential regulatory intervention that would help financial institutions fend off predatoryshort selling attacks is a requirement for financial institutions to hold more equity. Sucha requirement would make it less likely that the financial institution enters the vulnera-bility region and thus reduces the financial fragility that can be exploited by short sellers.Alternatively (if additional equity capital is costly), a requirement to issue at least someamount of reverse convertible debt that converts into equity if the leverage constraintbinds would allow the financial institution to reduce leverage without selling any of itslong-term asset holdings.

Second, in our model a credible promise to recapitalize the financial institution couldeliminate that “bad” equilibrium: If short sellers anticipate that the financial institution

18 See SEC release 34-58785, available at http://www.sec.gov/rules/final/2008/34-58785.pdf.19 See “Implementing aspects of the Financial Services Act 2010” available athttp://www.fsa.gov.uk/pubs/cp/cp1 18.pdf20 See “Europe’s naked short selling ban leaves investors with skin in the game,” Reuters,December 4, 2012.

Page 25: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

24 M. K. BRUNNERMEIER AND M. OEHMKE

never has to liquidate any of its long-term asset holdings, they will never attack, know-ing that they cannot break even on their short positions. However, such an interventioneffectively amounts to a government guarantee for the financial sector. In a richer modelwhere financial institutions choose their investments, this would lead to substantial moralhazard concerns. It is thus not clear that such a guarantee would be desirable (at least ifit is anticipated ex ante).

Third, to the extent that regulators use temporary short-sale bans to protect vulnerablefinancial institutions from predatory short selling attacks, this will likely increase theimportance of prompt corrective action by regulators. Specifically, if short-sale bans makeit harder for market participants to single out financial institutions that should be shutdown or restructured (e.g., so-called zombie banks), then the regulator’s role in identifyingthese institutions becomes all the more important.

Panic sales. While the focus of this paper is on short selling, another novel result ofour analysis is that, in addition to short sales, regulators may also want to consider thepossibility of destabilizing “panic sales” by current investors in the financial institution’sequity. In particular, if current shareholders fear that other shareholders are likely to selltheir holdings, such panic sales can become self-fulfilling for financial institutions that aresubject to leverage constraints of the type described in our model. Hence, akin to theclassic bank run problem described in Diamond and Dybvig (1983), our analysis impliesthat regulators may want to have an eye on financial market runs on the equity of forfinancial institutions that are subject to leverage constraints. As shown in Proposition 3,this concern is stronger the less likely it is that current shareholder can coordinate theiractions.

4.2 EMPIRICAL PREDICTIONS

Existing empirical evidence. A number of recent papers empirically examine the effectsof recent short selling bans in the U.S. and Europe. For example, Beber and Pagano (2013)use international data to document the effects of short sale bans across different markets.Boehmer, Jones, and Zhang (2013) provide a detailed investigation of the U.S. ban onshort sales imposed by the SEC in September 2008. The main findings of these (and anumber of other) studies is that short sale bans led to an unambiguous and significantreduction in liquidity, market quality, and the speed of price discovery (as measured, forexample, by bid-ask spreads, effective spreads, or price impact measures). However, theresults regarding the effect of short sale bans on prices—perhaps the main motivationfor intervention—are much weaker. For example, Boehmer, Jones, and Zhang (2013) findsignificant abnormal excess returns only for the largest U.S. financial institutions and pointout that those may have been caused by the Troubled Asset Relief Program (TARP), whichwas launched more or less at the same time. In fact, because their analysis suggests thatthe price effects were permanent, they conclude that the TARP may have been the morelikely cause.

Cross-sectional predictions from our model. Our analysis may help interpretsome of the findings of these recent empirical studies. Moreover, the empirical implica-tions of our analysis could potentially be helpful in the design of future empirical studiesof short selling bans on financial institutions. First, our model suggests that financial in-stitutions, as opposed to other firms, should be particularly affected by short sale bans.Second, the ability of short sellers to prey on financial institutions depends crucially on

Page 26: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 25

the financial condition of the financial institution. Hence in assessing the price effects ofshort-sale restrictions, one should control for leverage, maturity mismatch, or similar vari-ables that measure financial fragility. The cross-sectional prediction of our model is that itis vulnerable financial firms for which the price effects of short sale bans are largest. Third,our model highlights the importance of taking into account the potential multiplicity ofequilibria when interpreting the empirical evidence. For example, if investors expect that,with some probability, there is a switch to the dominated equilibrium in which the bankgoes bankrupt, the elimination of the bad equilibrium through a short selling ban maylead to a permanent price effect. This can be the case even if the ban itself is temporary:Investors now anticipate that regulators may impose another ban should financial insti-tutions re-enter the vulnerability region. Hence, according to our analysis the permanentprice increase documented by Boehmer, Jones, and Zhang (2013) could also be attributedto a short-sale ban that eliminates the likelihood of the dominated equilibrium.

The funding of blockholders. In addition to the cross-sectional predictions above, anovel prediction of our model is that the vulnerability of a financial institution to predatoryshort selling depends not only on its own balance sheet but also on the balance sheets (orfunding conditions) of its large shareholders and other potential support buyers. Specif-ically, if large shareholders or other support buyers are important in fending off shortselling attacks, their ability to do so depends on their own funding liquidity, as proxiedby the parameter BMAX in Proposition 4. Hence, our model makes the prediction that,in the cross-section, financial institutions with less well-capitalized blockholders, are morevulnerable to predatory short selling attacks. Moreover, in the time series, predatory shortselling attacks are more likely to be successful when funding conditions for blockholdersare tight on average, for example during financial crises. Finally, our model predicts thatthe vulnerability of financial institutions to predatory short selling depends on the abilityof blockholders (and short sellers) to coordinate. Empirically, the number of blockholdersmay provide a proxy for their ability to coordinate.

Predictions on price skewness. In addition to cross-sectional predictions spelledout above, our model also predicts that large downward price movements can occur whena financial institution enters the vulnerability region or the doomed region. This meansthat, in our model, short selling increases negative skewness in equity prices, which isopposite to the prediction in Hong and Stein (2003), where banning short selling leads tonegative skewness. Consistent with our prediction, Bris, Goetzmann, and Zhu (2007) findevidence that there is significantly less negative skewness in markets in which short sellingis either not legal or not practiced. In addition, our model makes the testable predictionthat, absent short-sale restrictions, negative skewness should be observed particularly forfinancial institutions with weak balance sheets (i.e., those that approach the vulnerabilityregion or doomed region).

5. Conclusion

This paper provides a simple model of predatory short selling. Predatory short sellingoccurs when short sellers exploit financial weakness or liquidity problems of a financialinstitution. In our model, predatory short sales occur in equilibrium because the dropin equity valuation caused by short sellers leads non-insured depositors and short-termcreditors to withdraw funding from the financial institution. Because of this effectiveleverage constraint, short sales can force the financial institution to liquidate long-term

Page 27: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

26 M. K. BRUNNERMEIER AND M. OEHMKE

asset holdings at a discount. The resulting value reduction can allow short sellers to breakeven on their positions.

Our analysis shows that financial institutions can be vulnerable to attacks from preda-tory short sellers when their balance sheets are weak. For financial institutions that aresufficiently close to their leverage constraints. In the vulnerability region there are twostable equilibria. One of these stable equilibria does not involve short selling, while inthe other predatory equilibrium short sellers force a complete liquidation of the financialinstitution’s long-term asset holdings. In the doomed region there is a unique predatoryequilibrium in which the financial institution liquidates its entire long-term asset holdings.The doomed region is larger, the better potential short sellers are able to coordinate theiractions.

While our model does not develop a full welfare analysis of bans on short selling, thepossibility of predatory short selling in the vulnerability region and the doomed regionprovides a potential justification for temporary short sale restrictions for financial institu-tions in those regions, although the benefits of such restrictions have to be weighed againstthe cost of preventing short sellers from performing their beneficial role of eliminating po-tential overvaluation.

Appendix

Proof of Lemma 1: Absent the leverage constraint, the fundamental value of the financialinstitution’s equity is given by XR−D0, irrespective of the short sellers’ actions. Thisfollows immediately from the fact that, in this case, the financial institution never hasto liquidate early. Given the fixed fundamental value XR−D0, competition among shortsellers ensures that the share price is equal to the fundamental value XR−D0. To see this,consider first the case in which the intercept chosen by long-term investors is larger thanthe fundamental value of equity, i.e., P > XR−D0. In this case, short sellers take a shortposition S > 0 and, because short sellers make zero profits in equilibrium, the equilibriumshort position S is such that the share price is equal to fundamental, i.e., P = XR−D0.Now consider the case in which the intercept chosen by the long-term investors is belowfundamental value. Analogously to before, short sellers now take a long position (theyact, more generally, as arbitrageurs) to ensure that the equity if fairly priced. Finally,when the intercept chosen by the is equal to fundamental value, i.e., P = XR−D0, shortsellers do not take a position in equilibrium, i.e., S = 0. Because in all of these cases thefundamental value of equity is fixed at XR−D0, predatory short selling cannot occur. Inall three cases, the equilibrium condition (3) can thus be rewritten as P − λS = XR−D0.

Proof of Lemma 2: In the case that the constraint is violated, the result follows di-rectly from solving (7) for ∆X(S). Combining this with the fact that no liquidation occurswhen the constraint is satisfied and that the maximum amount that can be liquidated isX, yields the result.

Proof of Lemma 3: The result comes from the fact that, in equilibrium, a change ineither P or λ will be exactly offset by a corresponding change in the equilibrium level ofthe short position S, such that the equilibrium condition P = P is satisfied. Equilibriumprices and the equilibrium amount that has to be liquidated by the financial institutionthus remain unaffected.

Page 28: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 27

Proof of Proposition 1: We first analyze the case δ < γ, which is the case highlightedin the proposition. For completeness, we then also briefly discuss the cases δ > γ and δ = γ,which are not discussed in the main text.

Intuitively, when δ < γ, the P -curve is steeper than the P -curve, as depicted in thebottom panel of Figure 1. We now consider the three regions in the proposition in turn.First, we compute the region in which no predatory short selling can occur (the safetyregion). Short sellers cannot break even on a predatory short position if, after forcingthe financial institution to liquidate its entire long-term asset holdings, the fundamentalequity value at t = 2 still exceeds the stock price that forces this maximum liquidation att = 1. In this case, no matter how aggressive their selling, short sellers have to buy backat a higher price than they receive when shorting the stock at date t = 1.

To calculate the parameter region for which this is the case, assume that short sellerschoose a short position S such that the entire portfolio of the financial institution isliquidated, i.e., ∆X(S) = X. This requires that

(1− γ)D0 − γ(P − λS)

Rδ(1− γ)= X, (A1)

which yields

S =P

λ+

(1− γ)(δXR−D0)

λγ. (A2)

This strategy cannot be profitable when the stock price that forces the financial institutionto unwind all of its long-term asset holdings is smaller than the fundamental value of equityafter such a liquidation, i.e.,

XR−D0 − λS < δXR−D0. (A3)

Using (A2) to solve (A3) for R yields

R >D0

δX, (A4)

which is the expression defining the safety region in the proposition. When (A4) is satis-fied, the financial institution is well capitalized, and the unique equilibrium is one withoutpredatory short selling and P = P = XR−D0. Intuitively, in the safety region the liqui-dation value of the financial institution’s long-term assets is higher than the face value ofoutstanding debt.

Second, consider the region in which D0γX

≤ R ≤ D0δX

(vulnerability region). In this region,

the leverage constraint is not violated in the absence of predatory short selling, since D0XR

≤γ. This means that there is still an equilibrium in which no predatory short selling occursand P = P = XR−D0. However, now there is also an equilibrium in which short sellersforce the financial institution to liquidate its entire asset holdings. In this equilibrium, bythe zero profit condition, we have

P = XR−D0 − λS = max [δXR−D0, 0] = P, (A5)

where S ≥ S. Since we know that in this region δXR−D0 ≤ 0, the equilibrium price mustbe P = 0. In words, the financial institution has to liquidate all of its long-term assets,which are not sufficient to repay debt, and the equity value is zero. Both of these twoequilibria are stable. Finally, there is a third, unstable equilibrium, in which only part ofthe financial institution’s long-term asset holdings are unwound. Denote the amount of

Page 29: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

28 M. K. BRUNNERMEIER AND M. OEHMKE

short selling in the unstable equilibrium by S∗. In this equilibrium, ∆X(S∗) < X, suchthat we must have

P − λS∗ = XR−D0 − (1− δ)R∆X(S∗). (A6)

Substituting in for ∆X(S∗) from equation (8) and solving for S∗ yields

S∗ =P

λ− (1− γ)(D0 − δXR)

λ(γ − δ). (A7)

Using this expression for S∗, we can determine the price in the unstable equilibrium as

P = P − λS∗ =1− γ

γ − δ(D0 − δXR). (A8)

Substituting into (A2) yields that the amount the financial institution has to liquidate in

the unstable equilibrium is ∆X(S∗) = Xγ− D0

XRγ−δ

, as stated in the proposition.Third, consider the region in which the leverage constraint is violated even in the

absence of predatory short selling, D0XR

> γ. In this region, long-term investors set the

intercept P equal to the fundamental equity value in the absence of short selling: P =AR−D0 − (1− δ)R∆X(0), where ∆X(0) satisfies

D0 − δR∆X(0)

XR− (1− δ)R∆X(0)= γ, (A9)

which yields ∆X(0) = XD0XR

−γ

δ−γ(1−δ). When ∆X(0) = X

D0XR

−γ

δ−γ(1−δ)< X, partial liquidation is

possible and the equilibrium price is given by P = XR−D0 − (1− δ)XRD0XR

−γ

δ−γ(1−δ), which

simplifies to P = XR−D0 − 1−δδ−γ(1−δ)

[D0 − γXR], which is the expression in the propo-sition. In the presence of competitive short sellers, this is an unstable equilibrium: Aperturbation that leads to a small decline in the financial institution’s stock price triggersselling by short sellers and drives the stock price to zero, forcing a complete liquidationof the financial institution’s long-term asset holdings. This full liquidation outcome is theunique stable equilibrium: Short sellers force the financial institution to sell its entire assetholdings, ∆X = X. Because δXR−D0 ≤ 0, the equilibrium price must be P = 0. This isthe doomed region.

To complete the proof, note that in the doomed region, short sellers (or, more generally,arbitrageurs) cannot profitably act as support buyers by pushing the stock price up such

that the financial institution does not have to liquidate assets. Because at price P = XR−D0, the leverage constraint is violated, preventing liquidation by buying the stock requiresdriving up the price to a level that strictly exceeds the fundamental value XR−D0. Sinceabsent liquidation the value of equity is exactly XR−D0, this cannot be an equilibrium.

Now we briefly consider the case δ > γ. Intuitively, when δ > γ, the P -curve is less steepthan the P -curve, as depicted in the left panel of Figure 1. The first thing to note is thatnow the vulnerability region disappears, since D0

γX> D0

δX. In this case, as long as R > D0

γX,

the financial institution is well capitalized and no predatory short selling occurs. As aresult, liquidation only takes place when R < D0

γX. When R < D0

γX, the financial institution

has to liquidate ∆X(0) = XD0XR

−γ

δ−γ(1−δ), if it can satisfy the leverage constraint through

a partial liquidation (i.e., ∆X(0) < X). If ∆X(0) ≥ X, the financial institution has to

liquidate its entire asset holdings. Because the P -curve is less steep than the P -curve, short

Page 30: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 29

sellers cannot force a full liquidation if the financial institution can satisfy the constraintthrough a partial liquidation.

Finally, consider the knife-edge case when γ = δ. In this case, the P and P curves havethe same slope. This means that when R > D0

γXpredatory short selling cannot occur, while

when R < D0γX

, the financial institution is forced to liquidate all its holdings and P = 0.

When R = D0γX

, the equilibrium price can lie on any point on the interval [0,XR−D0].Proof of Proposition 2: We again focus on the more interesting case δ < γ, which is

the one depicted in Figure 4. The stable equilibria in the presence of short selling followdirectly from Proposition 1. The equilibrium prices when short selling is restricted aredetermined as follows. As before, assume that the long-term investors are rational, suchthat they correctly anticipate the t = 2 payoff P = XR−D0 − (1− δ)R∆X(0). As long asthe leverage constraint is not violated, D

XR< γ, the financial institution does not have to

liquidate (∆X(0) = 0) and P = XR−D0. When DXR

> γ, the financial institution has to

liquidate an amount ∆X(0) = XD0XR

−γ

δ−γ(1−δ), which can be derived as in (A9). When ∆X(0) =

XD0XR

−γ

δ−γ(1−δ)< X, partial liquidation is possible and the equilibrium price is given by P =

XR−D0 − (1− δ)XRD0XR

−γ

δ−γ(1−δ), which simplifies to P = XR−D0 − 1−δ

δ−γ(1−δ)[D0 − γXR].

When ∆X(0) = XD0XR

−γ

δ−γ(1−δ)≥ X, then the financial institution is forced to liquidate all its

long-term asset holdings at t = 1 and P = 0 even in the absence of short sellers. Solvingthe above expression for R shows that full liquidation is required whenever R ≤ D0

δX(1+γ).

Proof of Proposition 3: We consider the large short seller first. Recall that the largeshort seller moves first and chooses S ∈ [0, SMAX]. The large short seller’s trade is thenexecuted at P (S) = P − λS. Then, the competitive fringe moves and chooses SC . Theorders of the competitive fringe are executed at P (S + SC) = P − λ(S + SC).

Conjecture a no-liquidation equilibrium. If SMAX < S∗, the large short seller does nothave a profitable deviation: For all S ∈ [0, SMAX] we have P (S) < P , such that a shortsale is unprofitable for the large short seller. Hence, the short seller chooses S = 0 andthe no liquidation equilibrium is indeed an equilibrium: SC = 0 is a best response for thecompetitive fringe.

Now assume that SMAX > S∗. Now the short sale is strictly profitable for the largeshort seller, since for any S ∈ (S∗, SMAX] we have P (S) > P (S). Hence, the large shortseller chooses S > S∗. This makes it optimal for the competitive fringe to choose SC =S − S (such that the zero profit condition holds for the competitive fringe). Hence, whenSMAX > S∗ predatory short selling becomes the unique equilibrium. Inserting (A7) for S∗

and solving for R yields

R <D0

γX+

γ − δ

(1− δ)γXλSMAX. (A10)

Accordingly, the parameter region in which multiple equilibria are possible is given by R ∈[D0γX

+ γ−δ(1−δ)γX

λSMAX, D0δX

]. The region with a unique predatory short selling equilibrium

(doomed region) expands to R ∈[0, D0

γX+ γ−δ

(1−δ)γXλSMAX

).

Now consider the case of regular sellers. Recall that the large shareholder and thecompetitive fringe hold SMAX and SMAX

C shares, respectively, and that SMAX + SMAXC = S.

The large shareholder moves first and chooses S ∈ [0, SMAX]. This trade is executed at

Page 31: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

30 M. K. BRUNNERMEIER AND M. OEHMKE

P (S) = P − λS. Then the competitive fringe chooses SC ∈ [0, SMAXC ]. The orders of the

competitive fringe are executed at P (S + SC) = P − λ(S + SC).

First consider SMAX < S − S∗. In this case, irrespective of the large shareholder’s de-cision to sell, the competitive fringe can still cause a coordination failure by sellingSMAXC > S∗ shares (in such a coordination failure equilibrium they will always sell the

maximum amount). Anticipating this, the large shareholder will sell some or all of his

shares at price P = P − λS: S = argmaxS[P − λS

]+ (SMAX − S)P (SMAX

C + S)

However, when SMAX > S − S∗, the large shareholder’s decision not to sell rules outthe coordination failure equilibrium. Given that the large shareholder does not sell, thecompetitive fringe cannot profitably coordinate to sell because P (SMAX

C ) < P (SMAXC ). The

unique best response is thus SC = 0 and no liquidation is the unique equilibrium. Usingthe conjectured equilibrium condition P − λS = 0 we can rewrite the condition SMAX >S − S∗ as

SMAX >XR−D0

λ−[P

λ− (1− γ)(D0 − δXR)

λ(γ − δ)

]. (A11)

Inserting P = XR−D0 and simplifying yields

SMAX >(1− γ)(D0 − δXR)

λ(γ − δ). (A12)

Accordingly, the parameter region in which multiple equilibria are possible is given by

R ∈[

D0γX

, D0δX

− γ−δ(1−γ)δX

λSMAX

]. The region where no liquidation is the unique equilibrium

(safety region) expands to R ∈(

D0δX

− γ−δ(1−γ)δX

λSMAX,∞).

The analysis is similar in the case with random execution and limit orders. The maininsight is that limit orders effectively allow traders to condition on whether they areexecuted first or second. Hence, we can analyze these two cases in turn. When the largetrader moves first, the equilibrium is exactly as just discussed. This leaves us to describethe equilibrium when the competitive fringe moves first. Consider the short selling casefirst. The competitive fringe makes zero profit in equilibrium, such that both SC = 0 andSC = S are possible in equilibrium. If the competitive fringe chooses SC = 0, then thelarge short seller picks S = argmaxS[P − λS] if SMAX > S∗ and S = 0 otherwise. If the

competitive fringe picks SC = S, the large short seller chooses S = 0.Now consider the case of regular sellers when the competitive fringe moves first. If

SMAX > S − S∗ (or equivalently SMAXC < S∗), then no matter what the fringe does, it

is optimal for the large trader to set S = 0. But then the competitive fringe optimallyresponds by setting SC = 0 given that any SC ∈ (0, SMAX

C ] leads to strictly negative profits.Hence, the unique equilibrium is the no-liquidation equilibrium.

If SMAX > S − S∗, then there are multiple equilibria. In the “good” equilibrium, thecompetitive fringe chooses SC = 0 and the large trader optimally responds by settingS = 0. In the “bad” equilibrium, the competitive fringe sets SC = SMAX

C > S∗. Thelarge shareholder optimally responds with S = argmaxS[P − λ(SMAX

C + S)] + (SMAX −S)P (S + SMAX

C ) > 0. The optimal amount sold by the large shareholder is strictly lessthan SMAX, since this would lead to a zero payoff.

Taken together, the setup with random execution and limit orders thus leads to exactlythe same equilibrium regions as the setup in which the large trader moves first. The onedifference is that in the coordination failure equilibrium in which the large shareholder

Page 32: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 31

moves second he may sell less aggressively, internalizing that he will be executed at arelatively lower price.

Proof of Proposition 4: The proof follows along similar lines to the proof of Propo-sition 3. We assume that the large short seller and support buyer simultaneously sub-mit their orders and are executed first. The large short seller chooses a short positionS ∈

[0, SMAX

]. The support buyer chooses a buy order B ∈

[0, BMAX

]. The orders of

the large short seller and support buyer are executed at P (S −B) = P − λ(S −B). Af-ter the large traders have moved, the competitive fringe chooses SC and is executed atP (S −B + SC) = P − λ(S −B + SC).

First, assume that SMAX > BMAX. The predatory short selling equilibrium is the uniqueequilibrium when, starting from a conjectured no-liquidation equilibrium, the short sellercan effect a profitable short sale, irrespective of the action of the support buyer. WhenSMAX −BMAX < S∗, the support buyer’s best response to any short position S ∈ [0, SMAX]is to lean against the short seller, making the short sale unprofitable. Hence, no liquidationremains an equilibrium. When SMAX −BMAX > S∗, on the other hand, any S > S∗ +BMAX is a profitable deviation for the large short seller. In this case, no liquidation cannotbe an equilibrium. In the unique liquidation equilibrium, the support buyer optimallychooses B = 0 (for any B > 0 he would lose money) and, by the zero profit condition, the

competitive fringe chooses SC = S − S.Now consider the case SMAX ≤ BMAX. In this case, the support buyer can always

profitably counter short sales by the large short seller, such that the upper boundary ofthe region in which the unique equilibrium involves short selling and liquidation is given byR < D0

γX. Taken together, the two cases imply that the region with a unique short selling

equilibrium (doomed region) is given by R ∈[0, D0

γX+ γ−δ

(1−δ)γXλ[SMAX −BMAX

]+).

References

Allen, Franklin, and Douglas Gale, 1992, Stock-price manipulation, Review of FinancialStudies 5, 503–529.Allen, Franklin, and Gary Gorton, 1992, Stock price manipulation, market microstructureand asymmetric information, European Economic Review 36, 624–630.Amihud, Yakov, and Haim Mendelson, 1986, Asset pricing and the bid-ask spread, Journalof Financial Economics 17, 223 – 249.Autore, Don M., Randall S. Billingsley, and Tunde Kovacs, 2011, The 2008 short sale ban:Liquidity, dispersion of opinion, and the cross-section of returns of us financial stocks,Journal of Banking & Finance 35, 2252 – 2266.Beber, Alessandro, and Marco Pagano, 2013, Short-selling bans around the world: Evi-dence from the 2007-09 crisis, Journal of Finance 68, 343–381.Benabou, Roland, and Guy Laroque, 1992, Using privileged information to manipulatemarkets: Insiders, gurus, and credibility, Quarterly Journal of Economics 107, 921–958.Boehmer, Ekkehart, Charles M. Jones, and Xiaoyan Zhang, 2008, Which shorts are in-formed?, Journal of Finance 63, 491–527.

, 2013, Shackling short sellers: The 2008 shorting ban, Review of Financial Studies26, 1363–1400.Boehmer, Ekkehart, and Julie Wu, 2013, Short selling and the price discovery process,Review of Financial Studies 26, 287–322.

Page 33: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

32 M. K. BRUNNERMEIER AND M. OEHMKE

Bond, Philip, Alex Edmans, and Itay Golstein, 2012, The real effects of financial markets,Annual Review of Financial Economics 4, 1–22.Bris, Arturo, William N. Goetzmann, and Ning Zhu, 2007, Efficiency and the bear: Shortsales and markets around the world, Journal of Finance 62, 1029–1079.Brunnermeier, Markus K., 2005, Information leakage and market efficiency, Review ofFinancial Studies 18, 417–457.

, Gary Gorton, and Arvind Krishnamurthy, 2013, Liquidity mismatch, inMarkus K. Brunnermeier, and Arvind Krishnamurthy, ed.: Risk Topography: SystemicRisk and Macro Modeling (University of Chicago Press).Brunnermeier, Markus K., and Martin Oehmke, 2013, The maturity rat race, Journal ofFinance 68, 483–521.Brunnermeier, Markus K., and Lasse H. Pedersen, 2005, Predatory trading, The Journalof Finance 60, 1825–1863.Chakraborty, Debasish, and Bilge Yilmaz, 2004, Informed manipulation, Journal of Eco-nomic Theory 114, 132–152.Chang, Eric C., Joseph W. Cheng, and Yinghui Yu, 2007, Short-sales constraints and pricediscovery: Evidence from the hong kong market, The Journal of Finance 62, 2097–2121.Corsetti, Giancarlo, Amil Dasgupta, Stephen Morris, and Hyun Song Shin, 2004, Doesone soros make a difference? a theory of currency crises with large and small traders, TheReview of Economic Studies 71, 87–113.Corsetti, Giancarlo, Paolo Pesenti, and Nouriel Roubini, 2002, The role of large players incurrency crises, in Sebastian Edwards, and Jeffrey A. Frankel, ed.: Preventing CurrencyCrises in Emerging Markets (University of Chicago Press).Dechow, Patricia M, Amy P Hutton, Lisa Meulbroek, and Richard G Sloan, 2001, Short-sellers, fundamental analysis, and stock returns, Journal of Financial Economics 61, 77 –106.Desai, Hemang, Srinivasan Krishnamurthy, and Venkataraman Kumar, 2006, Do shortsellers target firms with poor earnings quality? evidence from earnings restatements, Re-view of Accounting Studies 11, 71–90.Diamond, Douglas W., and Philip H. Dybvig, 1983, Bank runs, deposit insurance, andliquidity, Journal of Political Economy 91, 401–419.Diamond, Douglas W., and Raghuram G. Rajan, 2001, Liquidity risk, liquidity creation,and financial fragility: A theory of banking, Journal of Political Economy 109, 287–327.Diamond, Douglas W., and Robert E. Verrecchia, 1987, Constraints on short-selling andasset price adjustment to private information, Journal of Financial Economics 18, 277–311.Farhi, Emmanuel, and Jean Tirole, 2012, Collective moral hazard, maturity mismatch,and systemic bailouts., American Economic Review 102, 60 – 93.Gerard, Bruno, and Vikram Nanda, 1993, Trading and manipulation around seasonedequity offerings, The Journal of Finance 48, 213–245.Goldstein, Itay, and Alexander Gumbel, 2008, Manipulation and the allocational role ofprices, Review of Economic Studies 75, 133–164.Goldstein, Itay, Emre Ozdenoren, and Kathy Yuan, 2013, Trading frenzies and their impacton real investment, Journal of Financial Economics 109, 566–582.Goldstein, Itay, and Ady Pauzner, 2005, Demand-deposit contracts and the probability ofbank runs, Journal of Finance 60, 1293–1327.Harris, Lawrence, Ethan Namvar, and Blake Phillips, 2013, Price inflation and wealthtransfer during the 2008 sec short-sale ban, Journal of Investment Management.

Page 34: Predatory Short Selling - Columbia Business SchoolPredatory Short Selling Markus K. Brunnermeier1 and Martin Oehmke2 1Princeton University, NBER, and CEPR; 2Columbia University Abstract.

PREDATORY SHORT SELLING 33

Harrison, J. Michael, and David M. Kreps, 1978, Speculative investor behavior in a stockmarket with heterogeneous expectations, Quarterly Journal of Economics 92, 323–336.Hong, Harrison, and Jeremy C. Stein, 2003, Differences of opinion, short-sales constraintsand market crashes, Review of Financial Studies 16, 487–525.Jacklin, Charles J., and Sudipto Bhattacharya, 1988, Distinguishing panics andinformation-based bank runs: Welfare and policy implications, Journal of Political Econ-omy 96, 568–592.Khanna, Naveen, and Richmond D. Mathews, 2012, Doing battle with short sellers: Theconflicted role of blockholders in bear raids, Journal of Financial Economics 106, 229–246.Khanna, N., and R. Sonti, 2004, Value creating stock manipulation: Feedback effect ofstock price on firm value, Journal of Financial Markets 7, 237–270.Kolasinski, Adam C., Adam V. Reed, and Jacob R. Thornock, 2013, Can short restrictionsacually increase informed short selling?, Financial Management 42, 155–181.Kumar, Praveen, and Duane J. Seppi, 1992, Futures manipulation with “cash settlement”,Journal of Finance 47, 1485–1502.Kyle, Albert S., 1985, Continuous auctions and insider trading, Econometrica 53, 1315–1335.Lioui, Abraham, 2011, Spillover effects of counter-cyclical market regulation: Evidencefrom the 2008 ban on short sales, Journal of Alternative Investments 13, 53–66.Liu, Xuewen, 2011, Short-selling attachs and creditor runs: When morris-shin meetsdiamond-dybvig, Working Paper, HKUST.Marsh, Ian W., and Richard Payne, 2012, Banning short sales and market quality: Theuk’s experience, Journal of Banking and Finance 37, 1975–1986.Miller, Edward M., 1977, Risk, uncertainty, and divergence of opinion, Journal of Finance32, 1151–1168.Ozdenoren, Emre, and Kathy Yuan, 2008, Feedback effects and asset prices, Journal ofFinance 63, 1939–1975.Postlewaite, Andrew, and Xavier Vives, 1987, Bank runs as an equilibrium phenomenon,Journal of Political Economy 95, 485–491.Saffi, Pedro, and Kari Sigurdsson, 2011, Price efficiency and short selling, Review of Fi-nancial Studies 24, 821–852.Scheinkman, Jose A., and Wei Xiong, 2003, Overconfidence and speculative bubbles, Jour-nal of Political Economy 111, 1183–1219.